Create 2d array from 2 single dimensional array in Python

Question

I have 2 single-dimensional NumPy arrays

a = np.array([0, 4])
b = np.array([3, 2])

I want to create a 2d array of numbers

c = np.array([[0,1,2], [4,5]])

I can also create this using a for loop

EDIT: Updating loop based on @jtwalters comments

c = np.zeros(b.shape[0], dtype=object)
for i in range(b.shape[0]):
    c[i] = np.arange(a[i], a[i]+b[i])

How can I achieve this via vectorization/broadcasting?

I agree with your first comment. But `np.arange(6).reshape(2,3)` won't work. 2 rows should be `np.arange(a[i], a[i]+b[i])` please note that `3` is not in my resulting array. — deepAgrawal, Jul 12 '21 at 07:26
Thanks for pointing that out. I have made edits based on @jwalters comments. — deepAgrawal, Jul 12 '21 at 15:03
When you come up with a 2d array target we can talk "vectorization". For what you currently want, lists will be just as good, maybe even faster. Look at your `c`! — hpaulj, Jul 12 '21 at 15:47

jtwalters · Accepted Answer · 2021-07-12T19:30:13.813

1

To create an ndarray from ragged nested sequences you must put dtype=object.

For example:

c = np.empty(b.shape[0], dtype=object)
for i in range(b.shape[0]):
    c[i] = np.arange(a[i], a[i]+b[i])

np.array([np.arange(a[i], a[i]+b[i]) for i in range(b.shape[0])], dtype=object)

Vectorized:

def func(a, b):
  return np.arange(a, a + b)

vfunc = np.vectorize(func, otypes=["object"])
vfunc([0, 4], [3, 2])

edited Jul 12 '21 at 19:30

answered Jul 12 '21 at 07:54

jtwalters

1

Is there a way to vectorize this? – deepAgrawal Jul 12 '21 at 15:00
@deepAgrawal Yes, I update the code with `np.vectorize` – jtwalters Jul 12 '21 at 19:30

1 Answers1